12-02-2012 09:25 PM
I guess my problem is a little complex here. I need to specify MISSING VALUES to be able to do IMPUTATION.
SoI have a dataset with several variables and each of the variables has missing values, however they are coded in some ways.
for example, data one looks like this,
actually, in vb1, "1" means missing; in vb2 "a" means messing and in vb3, "4" means missing and so on; Those values means "missing", "unknown", "unspecified", so they are missings just they've been coded.
I actually use proc sql to create a new table, and doing something like this,
input value $ value_d $;
2. use "proc sql" to combine tables together
But in the final results, sas looks at it like vb1 has "no missing values" with "3" (instead of 5) obsetrvations and same as for vb2, vb3...... in this way I have nothing missing since sas only looks at completed cases. I used proc means n nmiss options to get the results. so i don't know if my imputation is going to work well.
please let me know a quick way to set those values as missing values because my "several" variables are like.....30+, thank you all very much!!
12-03-2012 12:57 AM
Is the missing value different for every value?
If so there isn't much of a short cut. Formats may help speed it up though if you have some that are similar.
12-03-2012 03:11 PM
Could you show us exactly what your proc means step looks like? I'm wondering if you're using VB1 to VB3 as analysis variables or classification variables. Either way, you're not coding it exactly right to get the results you want.